Introduction

The central proposal of Optimality Theory is that phonological outputs are not derived by the interaction of ordered rules. Rather, outputs are freely generated and the actual output for any input within a particular language is the one which is optimal given the ranking of the relevant constraints in that language. While constraints are assumed to be universal, constraint rankings are language particular, so that interlinguistic variation may be accounted for by ranking the same constraints in different orders. To test this hypothesis, this study compares verbal reduplication in three Bantu languages: SiSwati, Kinande and Kikuyu. These languages were chosen because, as shown in section 2, in all three the reduplicant is realized as a two-syllable prefix to the verb stem. Not surprisingly, these similarities will be accounted for by proposing that the three languages share similar constraints on the shape and position of the reduplicant. More surprisingly, perhaps, differences in the realization of the reduplicant in these languages will also be accounted for by proposing that the languages share similar constraints. Variation results from ranking these constraints differently in the three languages.

While Optimality Theory gives an elegant account for some of the variations in the form of reduplicants in these languages, I will show that there are aspects of the reduplication patterns discussed which pose problems for some claims of the theory. In sections 3 and 4 I argue that these languages provide evidence for Springs (1990) proposal that nonmetrical prosodic constituents, as well as metrical ones, may define bases and templates for reduplication. This analysis thus challenges the proposal current since McCarthy and Prince (1986) that only metrical prosodic constituents play a role in Prosodic Morphology.